We’re on SAP and our invoices come in from everywhere — PDFs, scanned images, the occasional spreadsheet someone emails over. Manual entry is killing us. Is there a solid way to automate the extraction and push it straight into SAP without a human in the loop for every single document? Curious what setups people are actually running in production.
Been there — SAP invoice processing is one of those things that looks straightforward until you’re actually in it.
You’ve got two main directions to go. SAP-native stuff like SAP Intelligent RPA or the built-in OCR in some modules can work, but in my experience setup is a pain and they get cranky with anything that deviates from a standard format. The other route is a third-party document processing platform that integrates with SAP via standard APIs — pushes data into AP, procurement, or whatever custom fields you need.
We ended up going with Lido for this. It handles the extraction regardless of format, validates fields against our SAP master data, and flags anything weird for manual review before it touches the system. No templates needed, which was a big deal for us because our vendor base is all over the place format-wise. There are other tools in this space too — Hyperscience, Rossum, ABBYY — worth comparing depending on your volume and complexity.
The general workflow is pretty consistent across most of these: document comes in, key fields get extracted and matched to SAP transaction types, clean records auto-create, exceptions get queued for review. You’ll want to spend time upfront defining your validation rules — that’s where the real tuning happens.
Setup on our end took about three weeks, and we hit ROI faster than expected just from AP labor savings. Realistically budget 2-4 weeks for implementation and another month for exception handling to stabilize.
Jumping back to the accuracy discussion – man, we’ve definitely seen that here too! Do you guys find it really depends on the document type?
For us, invoices are usually pretty smooth sailing. Most tools seem to handle them without much fuss, and we get pretty solid extraction rates. It’s almost like they were built for invoices first, you know?
But man, purchase orders? Those are a whole different beast. Every single tool we’ve thrown at them – from the big names to some niche stuff – just seems to struggle. It’s like they can’t quite get a consistent read on all the variations, or maybe the way the data’s structured differently compared to an invoice throws them off. We spend way more time cleaning up PO data than anything else.
So yeah, just wondering if that’s a common experience out there, or if we’re just uniquely cursed with our POs! Would love to hear if others have the same headache.